diff --git a/docs/drivers/MTConnect.md b/docs/drivers/MTConnect.md
index 9999b3e4..42cf80ca 100644
--- a/docs/drivers/MTConnect.md
+++ b/docs/drivers/MTConnect.md
@@ -437,11 +437,11 @@ refused — advances `LastSuccessfulRead`, with the state following the read / `
 (a write cannot clear either of those paths' degradation). A write that got **no** answer
 (transport failure, deadline) sets **Degraded** with the stamp preserved and a type-only
 `LastError` (review 2026-10-21 I-3, coordinator ruling). One success writer is shared: that
-degradation is **latched until the next answered write or the next answered `/current` read**.
-Neither a `/sample` chunk nor a read served from the live pump index clears it (both are stream
-evidence, not an answer to a request), so on a **subscription-only** driver one timed-out write
-keeps the driver Degraded until the next write — the comms-loss watchdog does not trip on it,
-because the stream keeps advancing `LastSuccessfulRead`.
+degradation is cleared by the next answered **write**, the next answered **`/current` read**, or
+the next **`/sample` chunk** the current session's pump delivers (review r2 N-5: a delivered
+chunk proves the Agent is alive — the analogue of OpcUaClient's good keep-alive — so it cannot
+stick on a subscription-only driver). A read served from the live pump index does not clear it
+by itself; the chunks behind that index do.
 
 **One request per device (review 2026-10-21 T-6).** The slots that survive the refusals above
 go out as one `POST {agent}/{device}?id1=v1&id2=v2…` per device (the Agent's PUT surface takes
diff --git a/src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect/MTConnectDriver.cs b/src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect/MTConnectDriver.cs
index 816b877b..74e84efd 100644
--- a/src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect/MTConnectDriver.cs
+++ b/src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect/MTConnectDriver.cs
@@ -371,9 +371,10 @@ public sealed partial class MTConnectDriver
 
     /// <summary>
     ///     Whether the last Agent write got no reply (review I-3 / coordinator ruling: an unanswered write degrades).
-    ///     Latched until the next answered <c>/current</c> read or answered write (<see cref="RecordReadAnswered"/> /
-    ///     <see cref="RecordWriteAnswered"/>). <b>Neither a <c>/sample</c> chunk nor a live-index read clears it</b>, so on
-    ///     a subscription-only driver one timed-out write keeps the driver Degraded until the next write (r1 N-5).
+    ///     Cleared by the next answered <c>/current</c> read, answered write (<see cref="RecordReadAnswered"/> /
+    ///     <see cref="RecordWriteAnswered"/>) or delivered <c>/sample</c> chunk of the current session
+    ///     (<see cref="TryMarkStreamDelivered"/>, r2 N-5) — so it cannot stick on a subscription-only driver. A read served
+    ///     from the live index does not clear it directly; the chunks behind that index do.
     /// </summary>
     private volatile bool _writeDegraded;
 
@@ -856,9 +857,8 @@ public sealed partial class MTConnectDriver
 
         // The read path is answering (from a stream that is delivering), so a degradation a failed /current left
         // behind is no longer true of it. The stamp stays the pump's — this read heard nothing from the Agent itself.
-        // Clears only the READ degradation (r1 N-5): an index read is the stream's evidence, not an Agent answer to a
-        // request — the same evidence a /sample chunk carries, which by the I-3 ruling does not clear an unanswered
-        // write either.
+        // Clears only the READ degradation: an index read is derived from the stream's evidence — the chunks that fed the
+        // index are what clear an unanswered write (TryMarkStreamDelivered, r2 N-5).
         if (_readDegraded)
         {
             _readDegraded = false;
@@ -2751,6 +2751,11 @@ public sealed partial class MTConnectDriver
 
             _streamDegraded = false;
             _pumpLive = true;
+
+            // r2 N-5 (coordinator): a delivered chunk proves the Agent is alive — the MTConnect analogue of OpcUaClient's
+            // good keep-alive clearing a read/write Degraded — so it also clears an unanswered-write degradation, which
+            // otherwise stuck forever on a subscription-only driver. Generation-checked with the rest (above).
+            _writeDegraded = false;
             PublishHealthy(DateTime.UtcNow);
 
             return true;
diff --git a/tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect.Tests/MTConnectBatchedWriteTests.cs b/tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect.Tests/MTConnectBatchedWriteTests.cs
index c230abbb..8858fbe3 100644
--- a/tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect.Tests/MTConnectBatchedWriteTests.cs
+++ b/tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.MTConnect.Tests/MTConnectBatchedWriteTests.cs
@@ -167,11 +167,12 @@ public sealed class MTConnectBatchedWriteTests
     }
 
     /// <summary>
-    ///     r1 N-5: stream evidence does not clear an unanswered write — neither a delivered /sample chunk nor a read
-    ///     served from the live pump index. The next answered write does (the positive control).
+    ///     r2 N-5 (coordinator ruling): on a subscription-only driver an unanswered write degrades, and the next delivered
+    ///     /sample chunk restores Healthy (a delivered chunk proves the Agent is alive). Positive control: with no chunk
+    ///     it stays Degraded.
     /// </summary>
     [Fact]
-    public async Task Stream_evidence_does_not_clear_an_unanswered_write_but_an_answered_write_does()
+    public async Task A_delivered_chunk_clears_an_unanswered_write_and_without_one_it_stays_degraded()
     {
         var (driver, client) = await WritableAsync();
         try
@@ -181,20 +182,19 @@ public sealed class MTConnectBatchedWriteTests
             client.WriteFailure = new HttpRequestException("connection refused");
             await driver.WriteAsync([new WriteRequest(PosRaw, 9.5)], Ct);
             driver.GetHealth().State.ShouldBe(DriverState.Degraded);
+
+            // Positive control: nothing else happens — it stays Degraded (no read, no write, no chunk).
             var currentCalls = client.CurrentCallCount;
+            driver.GetHealth().State.ShouldBe(DriverState.Degraded);
+            client.CurrentCallCount.ShouldBe(currentCalls);
 
             await client.PumpAsync(MTConnectPumpGenerationChunks.Pos(113, 118, "2.5")).WaitAsync(TimeSpan.FromSeconds(30), Ct);
-            (await driver.ReadAsync([PosRaw], Ct)).Single().StatusCode.ShouldBe(Good);
-            client.CurrentCallCount.ShouldBe(currentCalls, "the read was served from the live index");
 
-            driver.GetHealth().State.ShouldBe(DriverState.Degraded, "stream evidence is not an answer to the write");
-
-            client.WriteFailure = null;
-            await driver.WriteAsync([new WriteRequest(PosRaw, 3.5)], Ct);
-            driver.GetHealth().State.ShouldBe(DriverState.Healthy);
+            driver.GetHealth().State.ShouldBe(DriverState.Healthy, "a delivered chunk proves the Agent is alive");
         }
         finally
         {
+            client.WriteFailure = null;
             await driver.ShutdownAsync(CancellationToken.None);
         }
     }